1690 S.
MILTON
THOMSEN
Results and Discussion
Oxidation of iMgUzOs.sa.-The oxidation of MgU20s.84
was found to consist of two steps: (a) a low temperature
oxidation a t 200-300" which gave a product with oxygen
composition slightly above MgUzO6, and ( b ) a second step
requiring 525-550" for a conveniently measurable reaction
rate. The final product of the principal oxidation step is the
diuranate, with no evidence being found for the formation
of an intermediate phase corresponding t o MgU~06.8~.At
these temperatures the oxidation rate was dependent upon
the oxygen pressure only a t very low gas pressures. Oxi-
dation of MgU~06.6~ required heating a t 700°, confirming
the conclusion that this compound is not involved in the
oxidation of MgUzO6 to the diuranate.
The two steps in the oxidation of MgUzO5.84 to MgUzOt
were found to follow different mechanisms. The initial
l d expected on the
step was diffusion controlled as ~ ~ o u be
basis of previous data which had shown the existence of a
solid solution in this composition range (MgU~O5.84to MgUz-
0 8 04). The experimentally determined curve for this step
is not shown on Fig. 11, but it was found to follow closely
the curve shown for the strontium diuranate system. The
second step followed the phase-boundary controlled reac-
tion curve (Fig. 11, curve A ) over the entire range plotted.
The experimental points obtained on three separate oxida-
tions at 550 and 573' are plotted along the theoretical curve
A. The results indicate that the M ~ U L OtoS MgUZO, oxi-
dation involves little, if any, solid solution, and that the
diffusion of oxygen through oxidized product is not a rate-
determining factor.
Oxidation of CaU?05.95.-The oxidation of CaUzOs 95
followed a pattern somewhat similar to that of the corre-
sponding magnesium compound in that two steps were
again evident. The low temperature oxidation process
gave a product containing slightly more than six oxygen
atoms per calcium atom. The second oxidation step to
the diuranate took place in the same temperature range as
I he corresponding magnesium compound reaction.
[CONTRIBUTIOS FROM T H E RADIOCORPORATION
High-silica Fluosilicic Acids : Specific Reactions and the Equilibrium with Silica
BY S.MILTON THOMSEN
High-silica fluosilicic acids have escaped general recognition. They are readily prepared with any desired amount of
"extra" silica up to 18% more than the composition represented by the formula H2SiFa. They show characteristic reactions
with sodium salts and with fluorides, and these reactions are convenient for detecting and estimating the extra silica. They
attack glass. They react with hydrated silica, and an equilibrium constant for this reaction has been approximately evalu-
ated
Introduction.-Analyses on samples of fluosilicic
acid, reported' in 1936, showed the silica component
to be present in proportions larger than required by
the formula HzSiFs. Solutions prepared by dissolv-
ing hydrated silica in commercial fluosilicic acid have
been used to produce low reflection films on glass,
b o t h by a v a p o r process,2 and by a liquid bath3
process.
This paper reports (1) preparation of high-silica
fluosilicic acids (HSF acids) of predetermined com-
positions; (2) data on two previously reported4
specific reactions of HSF acids; and (3) data on
the equilibrium between silica and fluosilicic acid.
Terrnino1ogg.-In this paper, the term fluosilicic acid is
extended to apply t o any homogeneous liquid reaction
product of silica, hydrogen fluoride and water. All such
products constitute the three-component fluosilicic acid
system.
_______
(1) I v a n Tannanaev, J . Gcn. Chcm. U.S. S. R., 6, 1430 (1936).
(2) I;. H. Nicoll, RCA Review, X,440 (1949).
S. hl. Thomsen, i b i d . , XII, 143 (1961).
(3)
(i) S. M , Thomsen THISJOURNAL, 73, 2798 (1950).
1701. 74
The first portion of this second step of the oxidation curve
followed the theoretical phase-boundary controlled reaction
curve fairly well, but the latter portions more nearly re-
sembled the diffusion controlled oxidation. These observa-
-
tions can be explained by the thermal decomposition data
which show a diphasic region to CaU206.86 and solid solu-
tion extending from this point to the diuranate. Oxidation
to the diuranate apparently takes pldce by consecutive
rcactions, the phase-boundary reaction predominating ill
the initial portions, and giving way to the diffusion con-
trolled process in the later stages.
Oxidation of SrUzOj,8o.-Partial oxidation was observed
a t temperatures as low as 0 " . At room temperature the
oxidation rate fell essentially to zero as the composition
reached SrU206.2. Heating caused further reaction, 15 ith
complete oxidation requiring several hours a t 550".
As was expected from the phase diagram, the curve for
the rate of oxidation of the strontium compound is much
more like curve B (diffusion controlled) than like curve A
(phase-boundary). The deviation between the experimen-
tal curve and curve B may be due t o the formation of the
intermediate oxide indicated by the thermal decomposition
data.
Acknowledgment.--We are indebted to Dr.
W, H. Zachariasen of the University of Chicago,
to Dr. Stanley Siege1 of the Chemistry Division,
and to Dr. M. H. Rfueller of the Metallurgy
Division of Argonne National Laboratory for
the X-ray data and their interpretation. We
also wish to thank Dr. Norman R. Davidson
of the California Institute of Technology for
carefully reading the manuscript and for mak-
ing a number of valuable suggestions and com-
ments.
CHICAGO, I L L I S O I S SEPTEMBER
RECEIVED 4, 1951
OF AMERICA,RCA LABORATORIES]
Two numbers are required to specify the composition
of a fluosilicic acid. Used for this purpose are the concen-
tration, denoted by C, and the silica number, denoted by
S. Thus, C is the concentration of the hydrogen fluoride
component, expressed as moles of HzSiFe per liter, and S is
the number of moles of silica component present for each 6
moles of hydrogen fluoride component. Both C and S are
computed from data obtained by acidimetric titration.
Fluosilicic acid of composition corresponding t o the for-
mula HPSiFB,has S = 1. The HSF acids have S > 1, and
the low-silica fluosilicic acids have S < 1. At the lower
extreme is hydrofluoric acid, a degenerate fluosilicic acid
with S = 0; a t 4he upper extreme are the silica-saturated
fluosilicic acids, having S approximately 1.18, but varying
with C.
The silica component present in low-silica fluosilicic acids,
and in an acid of S = 1, is termed primary silica. The
HSF acids contain, in addition to such primary silica. a
secondary silica component with its own characteristic
chemical behavior.
Because the composition of silica-saturated fluosilicic
acids approach closely that of H?SiF&F, (fluodisilicic acid)
for which S = 1.2, the secondary silica will be considered
to be present in this form. The equation for the reaction
of silica to form an HSF acid is then
4H+ +5SiFC + +
Si02 +~ [ S ~ F V S ~ F J ]2H20
' (1)
April 5, 1952 FLUOSILICIC
HIGH-SILICA
Appropriate modifiers are used with the term silica to
distinguish several meanings. Primary and secondary
silica are components of a liquid fluosilicic acid; hydrated
silica is the C.P. reagent (silicic acid) as purchased; and
precipitated silica is a solid phase formed and separated
from a fluosilicic acid.
Preparation of High-silica Fluosilicic Acids.-Any fluo-
silicic acid of predetermined composition can be prepared
by di3solving the required amOunt of assayed hydrated
silica in a measured volume of standardized hydrofluoric
acid. Another procedure is more convenient when many
solutions, in a variety of compositions, are to be prepared.
I t consists in mixing the required amounts of two liquids:
15.0 N hydrofluoric acid (reagent I), and silica-saturated
2.50 Mfluosilicic acid (reagent 11), and diluting if necessary.
To prepare reagent 11, commercial fluosilicic (2.5 to 2.7
h l ) is diluted to 2.50 M . Hydrated silica is added, and
allowed to saturate the solution. The reaction is rapid,
as indicated by a temperature rise of about 2' with a typical
raw material. Saturation is complete when permanent
turbidity results from the addition of a small portion of
"fresh" hydrated silica, to provide some finely divided
material. The saturated solution is filtered through paper.
Reagent I is made 15.0 N to have the same value of C as
reagent 11. A mixture made from a ml. of reagent I ( C =
2.50, S = 0) and b ml. of reagent I1 ( C = 2.50, S = 1.18),
will be a fluosilicic acid with C = 2.50, and with S as follows
S = 11.8b/(a + b)
The 2.50 molar acid is finally diluted with water to obtain
the desired C; this does not change S.
Method of Analysis.-Compositions of HSF acids were
determined by an acidimetric m e t h ~ din , ~which the sample
is titrated in the presence of powdered sodium fluoride, to two
successive end-points.
To a 5.00-ml. sample in a plastic dish are added 3 g. of
powdered C.P. sodium fluoride, 0.5 ml. of 0.1% brom thymol
blue and 1.0 ml. of 0.1% phenolphthalein. After being
stirred, the mixture is titrated with 3 M sodium hydroxide
to a greenish-blue color (pH 8). The mixture is then trans-
ferred to a borosilicate glass beaker, diluted with water to
200 ml., heated nearly to boiling, and titrated a t that tem-
perature to the phenolphthalein end-point (PH 10).
The two quantities, a , milliequivalents of sodium hydrox-
ide in the first titration, and b, total milliequivalents for both
titrations, characterize the sample. From them are com-
puted the two quantities C = b/6v ( u = ml. of sample) and
S = 3(b - a ) / 2 b (independent of sample volume), defined
above.
Precipitation of Secondary Silica.-Sodium chloride
solution is a useful reagent for distinguishing the HSF
acids from the low-silica fluosilicic acids, and will give an
indication of the amount of secondary silica present. The
primary silica separates as a sodium fluosilicate precipitate
immediately in either case, upon adding the reagent. Then
follows a slow separation of the secondary silica as a silica
precipitate, if present. Turbidity appears and increases
for perhaps a half-hour; after several hours the precipitated
silica has settled out. Two experiments will be described,
illustrating the iise of this reaction.
In the first experiment, enough sodium chloride solution
was added to each of four different samples of fluosilicic
acid to precipitate all the fluosilicate. These samples had
silica numbers of 0.94, 1.00, 1.06 and 1.12; all had C =
2.50. To 10 ml. of each was added 15 ml. of 4.0 M sodium
chloride solution. The sodium fluosilicate precipitates were
allowed to settle, and the clear, supernatant liquids were de-
canted and allowed to stand. The last two slowly became
turbid and after a few hours formed silica precipitates, the
fourth having a precipitate twice as voluminous as the
third. The first two remained clear. Thus, the primary
and secondary silica are separately precipitated.
In the second experiment, progressively greater amounts
of sodium chloride solution were added to samples of one
particular 2.50 Mfluosilicic acid. Into each of 13 test-tubes
was put a 10-ml. sample of the acid. To each was added 4.0
M sodium chloride solution, 1 ml. to the first, 2 ml. to the
second, thus up to 13 ml. to the 13th tube. The sodium
fluosilicate precipitate was allowed to settle, the clear liquids
were decanted into other tubes, and allowed to stand.
With silica-saturated fluosilicic acid (S = 1.18), all sam-
(5) S. hl. Thomsen, Anal. Chem., 23, 973 (1951).
ACIDS 1691
ples showed turbidity and silica precipitation. With
acid of S = 1.00, none showed silica turbidity. With
acid half-saturated with silica ( S = 1.09), the first six
samples remained clear, and the rest developed silica pre-
cipitates. In this third case, precipitation of half or less of
the fluosilicic acid as sodium fluosilicate, left enough to keep
all secondary silica in solution.
Progressive addition of sodium chloride solution to an
HSF acid, as it removes SiFr from solution, enriches the
solution relatively in the silica component (increases S ) up
to the saturation value. Thereafter a silica precipitate also
forms, though with the time delay already mentioned. The
equation may be written as
2H20 + 3[SiFc,SiF4]' + 10Na+ +
5Na2SiFe -1+ +
Sio2-1 4H + (2)
Neutralization of an HSF acid also results in the separa-
tion of the secondary silica, and this may be represented by
the pair of equations
40H- + 3SiFs.SiFT +5SiFT + Si02 + 2H20 (3)
60H + 6 H + +6H20 (4)
Thermal Effects .-The following thermal data, while
not obtained by accurate calorimetry, are informative and
useful.
The reaction of Eq. 1, occurring as hydrated silica dis-
solves to form secondary silica, is exothermic. The addi-
tion of excess hydrated silica to a fluosilicic acid of initial
composition C = 2.58, S = 1.024, caused a temperature
rise of 2.1". After saturation, the composition was C =
2.58, S = 1.185. Computation from these figures gives
approximately 5 kcal./mole S O 2 , for the heat evolved.
Addition of excess hydrated silica to 6 N hydrofluoric acid
gave a rise of 30' under the same conditions, or about 30
kcal./mole SiO2.
Heat is also evolved when a fluoride is added to an HSF
acid, thereby converting the secondary silica to primary
silica, according to the equation
[SiFe?3Fn]' + 2F- +2SiFT (5)
Addition of powdered sodium fluoride to a silica-sprated
2.50 M fluosilicic acid gave a temperature rise of 19 .
The mixing of equal volumes of 6 N hydrofluoric acid and
a 2.50 M HSF acid, results in a temperature rise which is a
linear function of S of the HSF acid, as shown in Fig. 1.
This effect can be used to make a rapid approximate deter-
mination of s. From the data shown in Fig. 1, a figure of
about 25 kcal. per mole of secondary silica is computed as
the heat evolved.
0.9 1.0 I. I 1.2
S I L I C A NUMBER.
Fig. 1.-Effect of the silica number (S)of 2.50 Mfluosilicic
acid on the temperature rise occurring when it is mixed with
an equal volume of 6 N hydrofluoric acid.
Thus, more heat per mole of silica is evolved (25 to 30
kcal.) when hydrated silica, or secondary silica of an HSF
acid, reacts with hydrofluoric acid to produce primary
silica (in the form of SiFr) than when hydrated silica re-
acts by Eq. 1 (5 kcal.) to form secondary silica. In the
latter case, Si-F bonds are presumably redistributed to in-
I692 S. MILTONTHOMSEN Vol. 74

volve more Si atoms, without an increase in the number of
bonds.
Behavior Toward Glass.--The long-held belief"' that a
product corresponding t o the formula H2SiF6will not attack
glass, is erroneous. It is apparently based upon the mis-
taken belief that the silica-saturated products produced by
the usual laboratory methods correspond t o this formula.
Fluosilicic acids not saturated with silica do attack glass.
Commercially available fluosilicic acids, generally in this
class, are not sold in glass containers. Six lots had silica
numbers of 1.03, 1.06, 1.18, 1.04, 1.10 and 1.01. The last
three were drums of acid bought some months apart from
the same manufacturer.
Even the silica-saturated fluosilicic acids attack ordinary
glass, such as window glass. A process for producing
low-reflection films3 on glass surfaces employs the slow,
selective attack of a fluosilicic acid bath supersaturated
with silica to the extent of 2 to 10 millimoles of silica per
liter.
In Pyrex bottles, however, silica-saturated fluosilicic
acids in concentrations from 0.5 to 2.5 31 remained un-
changed in composition after storage for a year, and the
bottles showed no evidence of attack.
Vapor from HSF acids also attacks ordinary glass. In
doing so, it produces low-reflection filnis2~8 whose properties
vary with the S of the acid. XYhile the rate of attack (depth
z1.s. time) is directly proportional to C, the intensity of at-
tack, as measured by the percentage of glass removed from
the filmed area (which in turn controls the indes of refrac-
tion of the film) diminishes cis S increases.
Evaluation of Published Data.-Data, such as
that on specific gravities of fluosilicic acid solu-
tions$ and that on compositions of vapor in
equilibrium with acid of various concentrations, lo
are ambiguous in the absence of data on the compo-
sition of the acid used. Furthermore, the corn-
mon practice of specifying composition in percent-
age by weight introduces further ambiguity. X
silica-saturated acid, computed from titration to be
:30cG, HsSiFs,will actually have 31% of solute.
1.20, I I I I ,
I tions of SIFT and SiFs,SiFh, respectively
The Equilibrium with Silica.-Fluosilicic acids
of six different concentrations were saturated with
silica a t about 23" and the resulting products were
analyzed. The results are given in Table I.
Figure 2 shows that S increases slowly with in-
creasing C.
Each entry in Table I is for a separately pre-
pared solution. Many more were made a t C =
2.50 than the typical ones listed. The variations
in the figures of one column represent the uncer-
tainty of the analytical method rather than dif-
ferences in composition. This is known because
( I ) repeat titmtions gave variations of the same
magnitude, and ( 2 ) these solutions were used in the
liquid-bath process3 of making low-reflection films
on glass, a process so sensitive to composition that
differences of 0..5 millimole of silica per liter are
easily detected.
TL4BLE1
THESILICA SUMBER (5') AS A FUNCTION OF THE CONCENTRA-
n o x (C) I N SILICA-SATURATED FLUOSILICICACIDS
<. .% I8 ,\f 2.30 .II 2 00 -11 1.40 .M 1.00 I\. 0.50 .VI
.C 1 19i 1.185 1.175 1.170 1.167 1.164
1.190 1.181 1.167 1.167 1.158
1,188 1,168
1.187
1 185
~~ . ~ ~- ~~~ __ -
1.197 1 186 1.178 1.169 1.167 1.163
Such reproducible equilibria, together with the
absence of Tyndall Effect in the HSF acids, shows
that the secondary silica is not present colloidally.
Assuming the reaction to be that of Eq. 1, a con-
stant for the equilibrium with solid silica is the
following, in which M' and M" are the concentra-
k' = - d P
J T j ;
Table I1 gives the computed concentrations of the
ionic species present in each of the six silica-satu-
rated fluosilicic acids. These were computed by
considering the changes occurring according to Eq.
I, as silica dissolves in fluosilicic acid of the compo-
sition corresponding to the formula H2SiF6, and of

I
concentration C. Initially, (Hf) = 2C, 111' = C,
M'' = 0, and S = 0. After the silica is dissolved,
and S has been evaluated by titration, the following
relations hold : moles/liter of silica dissolved =
C(S - 1) = .v; ;II" = 3x; M' = C - 5x; and
(H+)= 2 C - dx. Yo activity coefficients are
used, and complete ionization of the acids is as-
1.151 I I I I sumed.
0 I 2 3 4
C O N C E N T R A T I O N , M O L E S H2SiF6 PER L I l E R ( C ) .
TABLEI1
COMPUTATJON OF EQUILIBRIUM COXSTAST
Fig. 2.-Silicanumber (S) at equilibrium, as a function of
k (from first five values of K ) = 3 ?c 1
log '
fluosilicic acid concentration (Cj.
_____ C 0.50 1.00 1.40 2.00 2.60 4.18
(6) J. PI'. Friend and R. hf.Caven, "A Textbook of Inorganic Chem- S 1.163 1.167 1.169 1.178 1.186 1.197
istry,'' Charles Griffin and Co., London, 1917, p. 193. AI" 0.24 0.51 0.71 1.07 1.40 2.48
(7) PI'-. V. Sidgwick, "The Chemical Elements and Their Com- *M I 0.10 0.17 0.22 0.22 0.17 0.05
pounds," Oxford University Press, London, 1950, p. 615. (H+) 0.68 1.33 1.86 2.58 3.14 5.06
(8) S. M . Thomsen. unpublished data. K 7000 350 6000 50 200 80,000
(9) "Handbook of Chemistry and Physics," 32nd Edition, Chemical
Rubber Publishing Co., Cleveland, Ohio, 1950. p. 1646. The first five values give an average of 3 i= 1 for
(10) J. W. Mellor, "A Comprehensive Treatise on Inorganic and
Theoretical Chemistry." Vol. 6, Longmans, Green & Co.. I.ondon,
log K. Using the value l o 3for K in the equation
1925, p. 941. -AF = RT1n k'
 OF TETRAFLUORO-
April 5, 1952 VAPORPHASEDIMERIZATION AND TRIFLUOROCHLOROETHYLENE
1693

gives 4 kcal. for - A F , which is near the 5 kcal. reaction as silica dissolves in an HSF acid.
per mole of silica, measured for the -AH of N. J.
PRINCETON, RECEIVED
JUNE 27, 1951

[CONTRIBUTION FROM THE UNIVERSITY OF COLORADO]

The Kinetics of the Vapor Phase Dimerization of Tetrafluoroethylene and

Trifluorochloroethylene
BY J. R. LACHER,
G. W. TOMPKIN
AND J. D. PARK

Using pressures of less than one atmosphere, the rates of dimerization of tetrafluoroethylene, chlorotrifluoroethylene and
their interdimerization have been studied. The reactions are second order and homogeneous. The Arrhenius equations
which will reproduce the experimental data are: 16.5 X 1O1O e--26.299’RT, 3.53 X 1Olo e - - 2 6 3 M I R T and 8.54 X 1Oloe-z630a’RT
cc./mole sec. -I, respectively. The significance of the symmetry numbers in determining the “A-factors’’ is discussed in
terms of the structures of the molecules involved.

Introduction TABLE
I
The successful dimerization a t high pressures of DIMERIZATION
OF CHLOROTRIFLUOROETHYLENE
tetrafluoroethylene and chlorotrifluoroethylene has Initial
been carried O U ~ . The~ ~ interdimerization
~ of the Run TQc
pressure,
mm.
k
mm.
-X, sec.
lo7,-,

two olefins has also been reported.* In all cases a 13 304.5 619.1 1.0
saturated cyclobutane was obtained. A kinetic 18 346.7 502.9 4.7”
study of the reactions was undertaken a t pressures 17 347.6 525.9 5.1
of less than one atmosphere and the results are re- 6 393.4 590.3 19
ported here. 5 393.6 607.0 18
Experimental Details 4 393.6 105.0 19
The reaction flask consisted of a one-liter Pyrex vessel 3 393.6 112.0 20
fitted with a thermocouple well and capillary entrance lines. 1 394.5 381.8 18
In order t o test for the homogeneity of the reaction, a flask 2 394.9 245.8 19
packed with Pyrex tubing was used. A Bodenstein quartz 16 439.1 623.2 65‘
spiral served to measure pressures to about 0.1 mm. The
reaction flask was placed in an electric furnace consisting of 15 439.8 615.8 65
an aluminum block suitably insulated. Temperatures 14 440.4 619.5 66
could be controlled to 0.2’ and there were no temperature 8 507.5 257.3 300
gradients present in the furnace to this accuracy. The 9 507.6 256.7 300
pressure change which accompanies the reaction was used
to follow its rate, and infrared analysis of the product was 10 507.6 190.0 310
used t o test for side reactions. In case of the interdimeriza- 11 508.7 143.5 310
tion of tetrafluoroethylene and chlorotrifluoroethylene the 12 508.7 104.3 320
rate was followed by infrared absorption. Samples of 7 509.1 347.9 310
tetrafluoroethylene and chlorotrifluoroethylene were kindly
furnished by the “Kinetic” Chemicals Division of the du a Seated in thermocouple. Packed cell.
Pont Company and they were purified by methods previously
described The rate constants were also independent of the extent of
The rate of dimerization of chlorotrifluoroethylene was the reaction. This is shown in Fig. 1. The ratio of the
studied a t temperatures between 304.5 and 509.1’ and a t rate constant for each increment to the rate constant for the
pressures of 100 and 600 mm. roughly. The reaction was initial increment is plotted as an ordinate and the per cent.
found to be second order. Second order constants were ob- completion of reaction is the abscissa. The logarithms of
tained from the differential equation the rate constants so obtained when graphed as a function of
AP/At = -kP2
I
Here A P is the change in the pressure in mm. of the rnono-
mer which occurs in the time interval At seconds. For P 2
the product of the pressure a t the beginning of the time in-
terval and that at the end were used. The constants so
obtained are summarized in Table I. They are independent
of the initial pressure as is shown in runs 1 through 6. No
significant change occurs a t 394” when the pressure varies
from 105 t o 623 mm. Run 16 shows that a packed cell is
without effect. I n run 18 a thermocouple was sealed into
the reaction flask so that the gases came in contact with it.
A very slight cooling (0.2”) of about five seconds duration
was noted when the reactor was initially filled. No signifi-
cant temperature rise was observed during the reaction.
(1) This work was supported by Contract N60NR-231, Task Order
VI,with the Officeof Naval Research, United States Navy. Presented
before the Symposium o n Fluorine Chemistry, 120th Meeting of the
American Chemical Society, New York, N. Y., September, 1951. 0 20 40 60 80
(2) Jesse Harmon, U. S. Patent 2,404,374. Reaction, %.
(3) A. L. H e m e and R. P. Ruh, THIS JOURNAL, 69, 279 (1947). Fig. 1.-Ratio of second order rate constant a t any time to
(4) Paul L. Barrick, U. S. Patent 2,427,116,
( 5 ) J. R . Lacher, J. D. Park, el ai., THISJOURNAL, 71, 1330 (1949). initial rate constant, plotted as a function of per cent. reac-
( 6 ) J. R. Lacher, J. D. P a r k , e l al., ibid., 71, 1334 (1949). tion. Dimerization of chlorotrifluoroethylene.